CALD - Cellular Automata-based Laser Dynamics
This application simulates the dynamics of laser devices using a Cellular Automata-based discrete model. Individual-based models like Cellular Automata are very effective to carry out detailed simulations of complex systems in a broad range of fields of science and technology. This kind of models has been recently applied by the proponent to simulate one of the most paradigmatic complex systems of particular technological importance: laser systems. Our application uses this model to carry out simulations intended to understand the emergence of macroscopic behaviours in lasers, arising from the interaction of simple microscopic components, and to simulate specific optoelectronic devices of arbitrary shape. Present results using low resolution simulations have probed that the model reproduces qualitatively many laser properties. But CPU-intensive high resolution simulations are necessary in order to simulate specific real laser devices. In addition to study the emergence of macroscopic properties, these simulations are useful as a modeling tool (alternative to the standard approach based on differential equations) for situations such as lasers ruled by stiff differential equations (which are difficult or impossible to integrate), difficult boundary conditions, very small devices (for which the usual approximations may not be valid), etc.
The laser device is simulated by a two-dimensional grid whose state evolves in time according to specific local evolution rules. The total number of laser photons and the population inversion in the system are computed for each time step. The behaviour of the system depends on three parameters. In order to characterize the possible outcomes of a laser device depending on the values of the three parameters, extensive parametric sweeps are necessary.
A preliminary estimate of the needed computing power is as follows: the computing time for a complete parametric sweep with detailed resolution, for a standalone standard PC, would be of the order of 30 days for a two-dimensional model and of the order of 8 years for a more complete three-dimensional model.
The original version of the application was a sequential version to carry out parametric sweeps. The application ran the simulation of the whole system on one computer, for a particular value of the three system parameters. This was an application with independent parallelism, specifically suitable to be executed on a Desktop Grid or a Service Grid infrastructure.
The application has been successfully ported to a BOINC DG infrastructure by researchers at the University of Seville. The aim of the application porting by W-GRASS within the framework of the EDGeS project was to run it through the BOINC to EGEE bridge and this way significantly increase the number of processors that are able to participate in the computation. The application is started as a Desktop Grid application. The bridge can register as a powerful BOINC client and send work units to a specified virtual organisation of EGEE.
The implemented solution was successfully tested on the Westminster Local Desktop Grid and work units were also processed by the DG to EGEE bridge.
During the performance tests the application was executed with variant redundancy, but the minimum quorum was kept as one. This means that more computers did the same work unit, however, the BOINC server had to wait only for the first result to assimilate the work unit. A test case of 792 work units was executed with redundancy values of 1,2,3 and 4 several times.
The following graph illustrates the achiweved speed-up as the function of the redundancy applied. As the figure shows, the optimal redundancy is about three. In that case the server generates 792*3=2376 work unit instances that are very close to the amount of work units that can be handled by the Westminster Desktop Grid at the same time. The maximum speedup achieved for the test-case was therefore nearly 200 when compared to the sequential execution.